Political Parties and Network Formation

Transcription

1 DISCUSSION PAPER SERIES IZA DP No Politial Parties and Network Formation Topi Miettinen Panu Poutvaara January 006 Forshungsinstitut zur Zukunft der Arbeit Institute for the Study of Labor

2 Politial Parties and Network Formation Topi Miettinen University College London and University of Helsinki Panu Poutvaara University of Helsinki and IZA Bonn Disussion Paper No January 006 IZA P.O. Box Bonn Germany Phone: Fax: Any opinions expressed here are those of the author(s) and not those of the institute. Researh disseminated by IZA may inlude views on poliy, but the institute itself takes no institutional poliy positions. The Institute for the Study of Labor (IZA) in Bonn is a loal and virtual international researh enter and a plae of ommuniation between siene, politis and business. IZA is an independent nonprofit ompany supported by Deutshe Post World Net. The enter is assoiated with the University of Bonn and offers a stimulating researh environment through its researh networks, researh support, and visitors and dotoral programs. IZA engages in (i) original and internationally ompetitive researh in all fields of labor eonomis, (ii) development of poliy onepts, and (iii) dissemination of researh results and onepts to the interested publi. IZA Disussion Papers often represent preliminary work and are irulated to enourage disussion. Citation of suh a paper should aount for its provisional harater. A revised version may be available diretly from the author.

3 IZA Disussion Paper No January 006 ABSTRACT Politial Parties and Network Formation We argue that anti-orruption laws may provide an effiieny rationale for why politial parties should meddle in the distribution of politial nominations and government ontrats. Antiorruption laws forbid trade in spoils that politiians distribute. However, itizens may pay for gaining aess to politiians and, thereby, to beome potential andidates for nominations. Suh rent-seeking results in exessive network formation. Politial parties may redue wasteful network formation, thanks to their ability to enter into exlusive membership ontrats. This holds even though anti-orruption laws also bind politial parties. JEL Classifiation: D7, D85, L14 Keywords: politial parties, two-sided platforms, rent-seeking, network formation Corresponding author: Panu Poutvaara Department of Eonomis P.O. Box 17 (Arkadiankatu 7) FIN University of Helsinki Finland We thank for useful omments Essi Eerola, Henrik Jordahl, Katarina Keller, Klaus Kultti, Mikko Leppämäki, Mikael Priks, Olli Ropponen, Juuso Välimäki and the partiipants of the XVII Annual Meeting of Finnish Eonomists in Maarianhamina, PET 005 in Marseille and seminar at HECER. The finanial support of Yrjö Jahnsson Foundation is gratefully aknowledged.

4 1 Introdution Anti orruption laws forbid trading attrative nominations made by politiians. But even though interested itizens are not able to buy suh spoils, it pays o to be on good terms with the politiians. Citizens buy tikets to fund raising events and spend time with politiians to be remembered when nominations are made. If there are no restritions on to whom politiians an alloate jobs, itizens an gain by rubbing shoulders with several politiians. This is time onsuming both for itizens and for politiians and results in wasteful network formation. Here, politial parties an provide valuable servies. Politial parties are powerful gatekeepers in modern demoraies. Citizens looking for positions of trust, and politiians alloating these, annot belong to more than one party. As politial parties an require their politiian members to give the nominations to other party members, parties an redue wasteful network formation between itizens and politiians. On the other hand, there is an additional linking ost when parties are present: parties also need to build links to politiians. We ompare network formation with and without the role of the politial parties. We take as our starting point that politial parties exist and that politiians distribute non ideologial spoils, suh as nominations to positions of trust or hire ivil servants who do not make politial deisions. We ask whether the gatekeeping role of the politial parties improves e ieny in the distribution of suh spoils. We also ask who gains and who loses from the parties meddling in the network formation. Our argumentation above suggests that party membership should be orrelated with the value of the spoils that politiians distribute. Indeed, there is suh a link: In Figure 1, we show how the share of voters belonging to a politial party (in 1997 to 000) is related to the total ompensation osts for the publi setor (as a perentage of GDP). 1 If Austria is exluded, the orrelation between party membership and total ompensation is A weaker but positive relationship holds for party membership and the GDP share of the publi setor (taken from Statistis Finland 005): without Austria, the orrelation is 0.8. The ase of Austria in itself also lends support to our analysis. In Austria, politial parties are well represented in the publi administration with many rms and organizations having traditionally double heads one for the Soial Demorats, one for the Conservatives. The strong politiization of nominations has been assoiated with exeptionally high level of party membership. (Enylopedia of Austria (005)) In our model, the party has full ontrol of the network formation of its members. We fous on the networks used to alloate non ideologial spoils; for onreteness think of positions of trust in a muniipality. In addition to providing exlusivity, politial parties alloate eah joining member to a unique politiian belonging to the party. Politial parties rst pay politiians 1 We inlude those European ountries whih where established demoraies already before the fall of Communism, and for whih data exists on both: Austria, Belgium, Denmark, Finland, Frane, Germany, Ireland, Italy, the Netherlands, Portugal, Spain, Sweden, Switzerland, and the United Kingdom. The data is from Mair and van Biezen (001) and OECD (001). Austria is an outlier: there 17.7 perent of voters belong to a politial party, while the share in all other ountries is between 1.6 and 9.7 perent. The R Student Residual for Austria is 1.47, well above the 5 perent ritial value of 1.96.

5 Austria Party membership as % of voters Party membership as % of voters Total ompensation Total for for publi setor workers as % of as GDP % of GDP Figure 1: The size of the publi setor and party membership for joining and then sell links to politiians to rent seeking itizens. 3 In equilibrium, parties, politiians, and itizens have rational expetations about the network struture. Equilibrium pries are determined by these expetations and inter party ompetition. Also in the no party equilibrium, expetations about the equilibrium network struture are rational. The equilibrium pries are determined by ompetition between politiians on the one side and itizens on the other side of the market. Given pries, politiians and itizens hoose the number of links they sell and buy, respetively. The more time itizens or politiians spend rubbing shoulders, the higher the opportunity ost of the time spent doing so. Thus, linking osts are onvex and, in equilibrium, the pries of links are determined by marginal osts. We rst derive the equilibrium networking struture and the equilibrium payo s in the no party and in the party equilibrium. We show that a no party equilibrium always exists and that the existene of a party equilibrium depends on the parameters of the model. Finally, we study the welfare properties of the equilibria. 4 We nd that the net e et of intermediaries on soial surplus depends on networking osts, the value of lurative nominations, and the relative numbers of politiians, itizens and parties. Therefore, a welfare maximizing government may well nd it optimal to promote the involvement of politial parties in some ases, and atively disourage it in others. Notie that alloating one rent seeking itizen to eah politiian would minimize the ost 3 Note that payments may take form of servies, in addition to money. 4 We refer only to the dimension of alloating the non ideologial spoils. Indeed, the role of the politial parties in distributing positions of trust and governments jobs has varied both between ountries and over time. 3

6 of network formation. Hypothetially, parties have two ways of aomplishing this: either by selling upoming nominations within the party or by allowing in only one itizen per eah nominating politiian. Yet, so as to the selling poliy, anti orruption laws bind also politial parties. At the same time, pre ommitment not to aept additional members is prohibited sine this would violate equal rights to politial partiipation. 5 We also nd that politiians would be better o if politial parties were not involved in the distribution of rents while itizens may gain from their presene, despite the fat that parties harge membership fees from itizens and transfer money to politiians! Even so, a party equilibrium may be inentive ompatible in that no single politiian would nd it optimal not to link with the party, provided that others do. Our model has ommon features with several strands of literature. First, we suppose that for a itizen to reeive a spoil from a politiian, a onnetion must be established between the two. This relates the urrent paper to the literature on ooperative networks, pioneered by Jakson and Wolinsky (1996). However, in our model agents may trade in the right to ontrol network formulation. This di erentiates our model from Jakson and Wolinsky (1996) and the subsequent literature. Seond, in our model, the itizens ompete for rents initiated by politiians as in rentseeking and lobbying ontests literature (Tullok (1967, 1980), Bernheim and Whinston (1986), Baye, Kovenok and de Vries (1993), Grossman and Helpman (1994), Besley and Coate (001) and Helpman and Persson (001)). Yet, there are two major di erenes. In our model, links are ostly for both itizens and politiians and endogenous, and require mutual onsent. Payments are made in exhange for establishing links. In rent seeking and lobbying literature, links are exogenous and ostless, and payments are bids in an aution or in a ontest. The only previous ontribution that endogenizes the relationship between politiians and lobbyists is Felli and Merlo (006). Our approah is omplementary to theirs. Whereas they analyze ideologial lobbying, we analyze lobbying on non ideologial spoils, like nominations and government ontrats. Furthermore, Felli and Merlo (006) assume that the links are ostless. Throughout the analysis, we assume that anti orruption laws work 6. Therefore, we have impliitly in mind a modern demoray with relatively low level of orruption. Previous literature on ontests has already analyzed extensively the ase where anti orruption laws an be irumvented. Third, our model is related to two strands of intermediation literature, the middlemen literature (Rubinstein and Wolinsky (1987)) on the one hand, and the literature on two sided markets (Rohet and Tirole (003), Armstrong (004)) on the other hand. In our model, the intermediaries, or the platforms, are the politial parties. Our approah di ers from these earlier ontributions most importantly in that the intermediary plays a useful role by restriting the ativity between the two sides of its market. There are two features whih separate politis and politial parties from other areas of intermediation. The rst one onerns the legal framework. Anti orruption laws restrit the 5 We ould generate the motivation not to restrit network formation by assuming that some itizens want to join parties to seek for rents, while others for ideologial reasons. Those joining parties for ideologial reasons favor a no restrition poliy. 6 The inability of politiians and itizens to trade nominations when these arise ould result from outside monitoring or from there being a fration of honest itizens and politiians who would report asking or o ering bribes, provided that punishments for orruption are su iently high. 4

7 ability of politiians and itizens to enter even into informal ontrats on alloating projets in exhange of payments. In other areas of intermediation (like the market of goods and ideas) the two sides are typially allowed to enter into private ontrats on the underlying goods or servies, rather than just on linking together. The seond di erene is in the servies that the intermediary provides. In previous literature, intermediaries are used only if they provide additional servies in reduing searh osts between the two sides of the market, information revelation of the underlying good or eonomizing transation osts. We nd that politial parties need not provide any of these servies; the servie they provide is exlusivity one annot be a member of several politial parties. Fourth, our explanation omplements previous e ieny rationales for the prominent role of politial parties. For example, Alesina (1988) and Alesina and Spear (1988) nd that politial parties may redue poliy utuations, ompared with the ase in whih subsequent ohorts of ompeting politiians with di erent preferenes would engage in one shot eletoral ompetition. Caillaud and Tirole (00) show that politial parties may make up voters information de it by designing and endorsing eletoral platforms. These previous ontributions leave the puzzle of why intermediaries arise also in ases where they do not redue the time spent searhing, provide additional information, or solve various ommitment problems. For example, it is questionable as to what extent a politial party would provide new information when lling positions of trust or publi jobs in a small muniipality. Yet, even these positions and many other jobs are typially earmarked to di erent politial parties. Our explanation for the role of a politial party applies even in these ases. The paper is organized as follows. Setion presents the model. Setion 3 analyzes equilibrium payo s and linking in the absene of parties, and setion 4 in their presene. Setion 5 explores whether politiians prefer a network with or without parties, and whether an eventual party equilibrium is inentive ompatible. Setion 6 presents a welfare omparison, and setion 7 empirial evidene. Setion 8 onludes. Model There are n t agents of type t, and there are three types t fa;b;cg. Type A is alled a politiian, i=1;:::;n A, type B is alled a itizen, j =1;:::;n B and type C is alled a party boss, k =1;:::;n C. The politiian reeives a hane to distribute a projet with probability p. 7 Eah politiian is indi erent to whih itizen to give the projet. Also, eah itizen is indi erent as to whih politiian from whom she reeives the projet. Yet, for politiian i to be able to give the projet to itizen j, there has to be a diret link between them. Party bosses an onnet to itizens indiretly, through politiians, but they need diret links to politiians. Note that our framework allows analyzing politis at di erent levels. At the national level, party bosses would be leaders of the national parties. At the muniipal level, they would be loal leaders, and politiians ould then be, for example, members of the muniipal ounil. Whihever the level, politiians require having a diret aess to their party boss. We take the 7 All the results would remain the same also if the politiian would reeive a ertain surplus, in ase a projet is ful lled. 5

8 identities of party bosses as exogenous 8. Citizens an be interpreted as individuals interested in positions distributed by eleted politiians or they an be interpreted as representatives of orporations interested in publi projets, the alloation of whih depends on deisions made by politiians. The strength of a link between agents a and a 0 is denoted m aa 0. Obviously, the strength of the link from a to a 0 must equal the strength of the link from a 0 to a, thus, m aa 0= m a 0 a. There is an indiret link between a and a 0, when there is a third agent a 00 to whom both a and a 0 are linked. We denote an indiret link between a and a 0 by aa 0. Maintaining a diret link requires time, and may require other osts. A dereasing marginal produtivity in other ativities, or an inreasing marginal utility of leisure, implies that the marginal ost of time spent on networking is inreasing. Furthermore, networking with ompeting agents simultaneously may pose additional hallenges. For example, a politiian who seeks to extrat ontributions from ompeting ontrators may need to spend more time in onvining these of the bene ts of giving. To apture these features, we model the marginal osts of networking as inreasing. A ost for i of maintaining a total amount m of links is 1 m where is a positive linking ost parameter independent of a player s type. The total amount of links in a network without politial parties need not be an integer. Formally, the width of eah link is between zero and one, zero implying no link at all and one implying a omplete link. The width of a link an be thought to be proportional to the time spent in maintaining the link. The probability weight that eah link reeives when a politiian alloates projets to itizens is proportional to the width of the link. The width of potential links with politial parties, on the other hand, is restrited to be either zero or one. This re ets the di erenes between the formal relationships with politial parties and the informal relationships between itizens and politiians. We assume that the number of agents is su iently large that itizens never ndit attrative tomaintain full links to all politiians. 3 Networks without Politial Parties Assume that there are more itizens than politiians,0 <n A <n B. In this setion, we assume that the politial parties do not partiipate in the network formation between politiians and itizens looking for projets distributed by politiians. There are times more itizens than politiians, n B = n A ; f;3; :::g. 9 If there are several itizens linked to a politiian, we assume that the politiian alloates the projet randomly, so that the probability that eah itizen is hosen is proportional to the width of the link. Thus the probability of getting the 8 In a riher framework, we ould model overlapping generations of politiians, with all young politiians being of type A and one of the old politiians beoming the party boss in the seond. 9 This simpli ation allows us to solve the model expliitly. Otherwise, the model would have to be solved numerially. 6

9 projet from politiian i equals m ji p : PØ=1;:::;n B m Øi Citizens are able to pay politiians for networking even if the itizens annot pay for the projets. For example, itizens interested in nominations an give ampaign ontributions to andidates or volunteer their time. Expliitly requiring a politiian to assign a nomination in exhange for suh ontribution, on the other hand, would be onsidered bribing and not lobbying. A itizen has to pay politiian i a reward, r i, for maintaining a full link. If the link is only partial, then the reward is redued proportionally. Note that this is a gross prie, and it has to ompensate the politiian for her marginal ost of linking. In addition to paying politiians r i for maintaining the links, itizens have to pay their own linking osts. The expeted payo of a itizen j reads 10 X m pp ji s X i Ø=1;:::;n B m Øi i The politiian i s maximization problem is as follows m ji r i (X i m ji ) : max m i1 ;:::;m inb f X j m ij r i (X j m ij ) g resulting in rst order onditions r i ( X j m ij )=0 for j=1;:::;n B. We assume that eah itizen ats atomistially 11 taking as given the reward, r i, and the total amount of links that eah politiian has to itizens, m ib, and maximizes max m j1 ;:::;m jna f X i p m ji m ib s X i m ji r i (X i m ji ) g: Thus, the rst order onditions write ps m ib r i X i m ji =0 (1) for i=1;:::;n A. It is easy to see that eah politiian is indi erent so as to whih itizens are linked to her. She only ares about the aggregate amount of links to itizens, m ib. Also, as long as linking 10 This formulation oinides with Tullok (1980) with R =1. Yet, the deision variable is not the ost of e ort, but rather, the number of links and the ost of linking is not linear but onvex in the number of links. 11 The appendix shows how the equilibrium derived here orresponds to an equilibrium of a dynami game. Notie that the atomistiity implies that, in the party equilibrium of setion 4, there is no strategi own side membership externality e et (see Rohet and Tirole, 004) even if suh an e et is non strategially present in our model. 7

10 rewards and the aggregate amount of links, m ib, are equal aross politiians, eah itizen is indi erent so as to how to alloate the links between politiians. All that matters is her aggregate amount of links, m ja. Sine linking ost funtions are onvex and idential aross politiians and itizens, the number of links is equal aross politiians, on the one hand, and aross itizens, on the other hand. Equilibrium rewards are equal to the politiians marginal linking ost. Thus, itizens equate their own marginal linking ost and the equilibrium reward, P n A j=1 m ji+rba, N to the expeted rent from linking,, where m N AB is the total amount of links in the equilibrium from a politiian to itizens and rba N is the equilibrium reward. Furthermore, we denote by m N BA the total amount of links that a itizen has to politiians in equilibrium. This allows us to derive the no party equilibrium 1. We summarize it as proposition 1: Proposition 1 In the no party equilibrium, the total supply of links by eah politiian is r ps m N AB = ( +1) ; () the total demand for links by eah itizen is r ps m N BA= ( +1) ; (3) and the reward for links from itizen to politiian is r ps r N = ( +1) : (4) Proof. See appendix. It is now straightforward to alulate the equilibrium surpluses. These are summarized in proposition : ps m N AB Proposition The equilibrium payo for the politiian is º N A = ps ( +1) whih is inreasing in p;sand : The expeted equilibrium payo for the itizen is º N B = ps ( +1) (5) (6) whih is inreasing in p and s and dereasing in. 1 We use the supersript N for the equilibrium values of endogenous variables in the no party equilibrium. Similarly, supersript P is used for the party equilibrium in setion 4. 8

11 Proof. The equilibrium payo s follow from plugging the equilibrium demand, supply and reward of the previous proof into the payo funtions. The politiian bene ts from the sarity of politiians, whereas the opposite holds for the itizen. As the number of itizens per politiian inreases, the demand of links of eah itizen dereases but the demand per politiian inreases. Thus, due to marginal ost priing and onvex osts, the pro t of the politiian inreases. Yet, for the itizen the strength of the link per eah politiian dereases and, thus, so does the probability of getting the projet. Yet, the reward that needs to be paid is higher. Hene, the payo for the itizen dereases. Surprisingly, the ost parameter of maintaining links enters neither the politiian s surplus nor that of the itizen 13. This implies that the e ieny gains due to the redution of the ost of maintaining the links are wasted in additional network formation. Notie that for any parameter values of the model the total amount of links in the equilibrium is positive and both the itizen and the politiian reeive a positive surplus. This implies the following proposition that establishes the existene of a no party equilibrium. Proposition 3 For any feasible parameter values of the model, there exists a no party equilibrium. 4 Networks with Politial Parties In this setion, we introdue politial parties as intermediaries that join together politiians and itizens. The servie that the parties provide turns out to be the exlusivity of links: no politiian or itizen an link to a member of another party. This redues the wasteful multipliation of links to eah politiian. Yet, there are osts to this as well, sine eah politiian must now link to the politial party in addition to the itizens. A party boss, whose type is denoted by C, exerts ontrol rights over his party (or its loal unit in the muniipality interpretation). He maximizes the party s surplus, net of his own linking osts. 14 Assume that there are politiians per eah party boss, where f;3;:::g Therefore, n A = n C and as n B = n A, n B = n C. We make the following assumptions: Assumption 1. The party boss reeives the right to ontrol and design the network of all the politiians and itizens linked with him on the ondition that the party bears all the linking osts. 15 Assumption. Politial parties, represented by party bosses, sign exlusive ontrats that state that itizens linked to them are not able to sign up with other politial parties. The politial parties annot ommit not to sign ontrats with additional politiians and itizens. 13 This result is not in the ore of our analysis. It may be due to the funtional forms and may not be robust to other spei ations. 14 We do not take a stane whether party bosses would keep the surplus, or part of it, for private onsumption, or if they use the surplus for ideologial purposes. 15 If the party would not bear the linking osts, then the party boss would have an inentive to require politiians to build more links ex post than they have agreed on ex ante. 9

12 Assumption 3. When itizens make their linking deisions, they know how many politiians belong to eah party. Assumption 4. Keeping up the links between politiians and itizens requires e ort or resoures, like in the ase with no politial parties. Also, links between party bosses and politiians require maintaining. Assumption 5. The party bosses play the ative role, making take it or leave it o ers to the politiians and itizens. Assumption 6. There must be a diret link between the politiian and the itizen who arries out the projet initiated by the politiian. 4.1 Properties of Party Equilibrium In this subsetion, we fous on an equilibrium where all politiians and itizens are party members every politiian and itizen has a diret or an indiret link to a party. In the next subsetion, we establish onditions for the existene of suh an equilibrium when eah politiian and eah itizen must ollet a non negative payo for joining a party. This provides an individual rationality ondition with a zero outside option. A positive payo outside option for the politiian is onsidered in setion 5.. There we study whether a politiian has an inentive not to join a party and link to itizens without the ontrol of the party. 16 For notational simpliity, the number of links that an agent of type t has to t 0 types is denoted by the same variable for all agents of the same type. This is restritive in general but, as in the no party equilibrium, it turns out to be a property of any equilibrium: all agents of the same type have an equal number of links and pay and reeive equal payments. There will be no diret payments between politiians and itizens sine the party regulates all links. 17 Yet, by assumption 6, the struture of the network within a party always inludes diret links between politiians and itizens. We denote the reward that eah itizen pays to the party boss by r BC and the reward that eah politiian gets from the party boss by r CA. A payment is made independently of the type of the link but the ost of a link is born only from diret links. We show in the next subsetion that both r BC and r CA are positive when the individual rationality onstraints are satis ed. 16 We do not onsider equilibria in whih the value of the projet is so low that not all itizens are willing to link to a party even if all politiians are party members. 17 Note that itizens often pay the party in the form of volunteer work. Our framework ould be generalized to allow for this, without hanging the qualitative results. Then the interpretation would be the following. Eah itizen pays to the party in the form of work up to the point in whih the marginal ost for itizen equals the marginal bene t for the party. The di erene between the monetarized value of this e ient work e ort and the equilibrium payment is settled in money. Party bosses may then let itizens (or part of them) work diretly for the politiian, thus transferring part of the ompensation in kind to the politiian. 10

13 The party s pro ts are º C (m CA ;m CB )=(m CB + CB )[r BC (m BA+m BC ) ] (7) (m CA + CA )[r CA + (m AB+ m AC ) ] (m CA+m CB ) : The rst term is the sum of rewards paid by the itizens to the party net of the itizens linking osts paid by the party. The seond term omprises the rewards to politiians and their linking osts paid by the party. The third term onsists of the party s own linking osts. By assumption, there must be a diret link between the politiian and the itizen who arries out the projet. This being the ase, the optimal struture of the network is suh that the party boss is diretly linked to the politiians and eah politiian is linked diretly to itizens. There are no diret links between the party boss and the itizens. Thus, (7) redues to º C (m CA ;m CB )= CB [r BC ] Equally, the itizen s expeted payo is m CA [r CA + (m AB+1) ] (m CA) : º B = p B s r BC : The probability that a given link results in a projet is Finally, a politiian linked to a party reeives a surplus equalling p B = p m CA m CB : (8) º A = r CA : In the party equilibrium, eah itizen and eah politiian is linked to a party. The next propositions 4 to 6 haraterize the party equilibria Proposition 4 In any party equilibrium, eah party boss is linked to politiians and itizens. The number of diret links are m P AB=, m P BA=1,m P AC=1;m P CA=, m P BC=0=m P CB. Proof. In the appendix. In addition, we an derive the equilibrium rewards paid by the party and the itizen and the equilibrium probability that a itizen gets the projet. 11

14 Proposition 5 In any party equilibrium, eah party pays a reward for eah politiian r P CA = ( + ) and eah itizen pays the party a reward r P BC (9) = ( +) (10) for the right to link to a politiian. The equilibrium probability of getting the projet is p P B= p : Proof. In the appendix. The equilibrium network struture is based on two priniples. First, the party is fored to build diret links between the itizens and the politiians. This being the ase, it is less ostly to the party boss to build his own links to the itizens via a politiian rather than diretly. Moreover, alloating an equal number of itizens to eah politiian minimizes the ost of linking. Seond, ompetition drives the bene t from an additional link equal to its marginal ost. Having a unique market reward and unequal number of links would violate the ondition of zero marginal net bene t. The one with less links ould apply the heapest network struture desribed above and get the same reward with a lower marginal ost. Hene, the number of links for any two agents of the same kind in the market must be the same. The rewards are suh that the parties are indi erent between selling an additional link to a itizen, or buying an additional link to a politiian, and stiking to the equilibrium number of links. Due to the marginal ost priing and the fat that eah politiian is alloated an equal number of itizens, the reward paid by the itizen to the party, (10), inreases in, the number of itizens per politiian, and in ; the linking ost parameter. Again, due to marginal ost priing, the reward that a party pays to the politiian equals the rewards from itizens linked to the marginal politiian, less the marginal politiian s osts of linking to the itizens and the linking ost of those itizens to the politiian, less the party s marginal ost of linking to the politiians. The surplus per politiian, rbc P ( +1), inreases in and in : On the other hand, the party s marginal ost of linking to the politiians inreases in and sine eah party links to politiians in equilibrium. Hene, the equilibrium reward to politiian dereases in and inreases in. As both the gross surplus per politiian and the marginal ost of linking to politiians inrease in, its net e et on the reward that politial parties pay politiians is unertain, a priori. Next, we haraterize the equilibrium payo s. Proposition 6 In any party equilibrium, the equilibrium payo s are º P C= ( +1) (11) º P A= ( + ) (1) º P B= ps ( +): (13) 1

15 Proof. These follow from inserting the equilibrium demands and rewards into the surplus expressions. The omparative statis of these equilibrium payo s are straightforward: Proposition 7 The omparative statis of the equilibrium payo s P > 0; > P P P P < 0 <0: The following orollary identi es ases where the itizen s party equilibrium payo is higher than her no party equilibrium payo : Corollary 1 The itizen s party equilibrium payo is higher than her no party equilibrium payo if ps ( +)( +1) : (14) ( +1) Above we found two ountervailing e ets of the linking ost on the party s equilibrium payo : the gross surplus per politiian inreases in and the marginal linking ost to politiians inreases in. Proposition 7 shows that the gross surplus per politiian e et always dominates. Proposition 7 reveals a on it of interest among the agents onerning the ost parameter. Whereas politiians and politial parties atually prefer a higher ost of networking, itizens prefer a lower one. That politiians and parties bene t from an inreasing transation ost follows from ompetition between politiians. When selling links to the itizens, the politial parties harge a reward, r P BC, equal to the marginal ost of adding one additional link. A lower value of the ost parameter would result in lower rewards and thus in lower equilibrium payo s to the parties and politiians Existene with Individual Rationality Constraints Let us now assume that eah politiian and itizen an refrain from joining a party, in whih ase they would reeive a payo of zero. The ondition for the existene of a party equilibrium is then haraterized in proposition 8: Proposition 8 There is a party equilibrium with º P A 0 and º P A 0 where eah itizen and eah politiian links to a unique party if and only if ps ( +) ++ (15) 18 This suggests that parties and politiians might objet to tehnologial innovations that might redue the osts of networking suh as politial partiipation via the internet. 13

16 that is if and only if the number of itizens per politiian is not too small or too large, the linking ost and the number of politiians per party boss is su iently small, and the expeted rent is su iently large. Proof. These results follow from the assumption that r CA 0 and from the requirement that º P B 0. The positive payo onstraints reate bounds for the number of itizens per politiian. The ondition + +is a requirement that when a politiian links with a party, the reward that the party pays to the politiian rca= P ( + ) ; (16) is not negative. This holds as long as the number of itizens per politiian is su iently high relative to the number of politiians per party. A positive reward is in line with what we observe in politis: politial parties typially pay their politiians in the form of ampaign ontributions et. rather than the other way round, and members pay parties membership fees rather than parties paying members. The equilibrium payo of the itizen, (13), is non negative if and only if the rst inequality in (15) holds. Inreasing the number of itizens per politiian su iently makes a itizen s expeted surplus negative. Combining the positive payo onditions with (14), we notie that it is possible that the itizen prefers the no party equilibrium to the party equilibrium or vie versa. Another point of interest, whih we report as a orollary to proposition 8 is that, in the party equilibrium, the number of itizens per politiian is smaller whenever the party equilibrium exists. Corollary If the party equilibrium exists, then the number of itizens linked to eah politiian is smaller in the party equilibrium than in the no party equilibrium. Proof. In the appendix. Asintheeonomyasawhole,alsoineahparty,thereare times more itizens than politiians by proposition 4. Thus, if the party let its politiians and itizens freely interat with the only restrition that none of its politiians or itizens ould interat with non members, the equilibrium payo s would not be a eted vis à vis the no party equilibrium. Yet, the party enfores the network design within the party so that eah itizen is linked to one politiian only. The total amount of links per politiian in the party equilibrium is always smaller than the total amount of links per politiian in the no party equilibrium, q ps ( +1). 5 Preferene for Party Membership In the equilibrium derived above, the party provides link exlusivity and network design servies for the itizens and the politiians. The individual rationality onstraints guarantee that eah politiian and eah itizen rather links to the party than remains inative. Yet, on the one hand, we do not know whether the politiians prefer the no party equilibrium to the party 14

17 equilibrium. If they did, they might be able to make a olletive binding deision not to link with any politial parties and gain. On the other hand, given that all the other politiians are linked to a party, eah politiian might prefer not to link with a party and build up her own network instead. We all a party equilibrium where the politiian s payo is higher than the outside option of staying out of any party and reating own network instead an inentive ompatible party equilibrium. In this setion, we study whether the politiians prefer to link diretly to itizens. We rst study the pro tability of a oordinated and olletive shift by the politiians from the party equilibrium to the no party equilibrium. Seond, we study the pro tability of a unilateral deviation of a single politiian from linking with a party to linking with itizens diretly when all other politiians remain linked with the politial parties. Notie that if a deviation by a single politiian from the party equilibrium pays o but a olletive deviation does not, then, from politiians perspetive, the struture is similar to that of a prisoner s dilemma. We show rst that politiians always prefer the no party equilibrium to the party equilibrium, and seond that with some parameter values, a deviation by a single politiian pays o whereas with others it does not. Thus, a dilemma struture never arises. 5.1 Politiians Preferenes Conerning the Equilibria We next identify whih of the equilibria is the politiians preferred equilibrium. Politiians prefer the no party equilibrium to the party equilibrium if the surplus in the former, (5), is greater than the surplus in the latter, (1). Hene, the ondition that eah politiian prefers the no party equilibrium is ps ( +1) ( + ): (17) When (17) holds, the no party equilibrium is preferred by the politiians. The next proposition shows that the politiians prefer the no party equilibrium. Proposition 9 The politiians prefer the no party equilibrium. Proof. In the appendix. This result is not surprising in the light of orollary. The total amount of links per politiian in the party equilibrium is always smaller than the total amount of links per politiian in the no party equilibrium, q ps. But, when ompeting for the politiians, the ( +1) parties pay politiians a reward equalling the marginal bene t of that politiian to the party whih on the other hand is just times the pro t that it makes per itizen that beomes a member of the party. This latter, on the other hand, is smaller than what the politiian makes per itizen in the no party equilibrium sine, as shown above, the total amount of links per politiian is smaller in the party equilibrium. 5. Inentive Compatible Party Equilibrium If a politiian annot be prevented from not linking with a party and linking with itizens diretly, the equilibrium surplus for a politiian has to be greater than or equal to the surplus 15

18 of the politiian linking diretly with itizens. Studying a politiian s inentives to remain linked with a party when all other politiians are linked with a party is the fous of this subsetion. We assume that a deviating politiian is able to make take it or leave it reward o ers to itizens potentially willing to link. The politiian is unable to ommit not to sell additional links. Therefore, she sells links until an additional link would give her a negative payo. This determines the number of itizens that will link to the politiian. In equilibrium, the itizens antiipate the total amount of links that a deviating politiian would like to sell. We show that there are ases where the politiian prefers to deviate and others where she does not. Proposition 10 There are feasible parameter values for whih the party equilibrium is inentive ompatible and others for whih it is not. Proof. The proof is in the appendix. 6 Welfare The aggregate surplus in the no party equilibrium, de ned as W N, equals W N 1 ps = n A ps n B ( +1) n 1 ps A ( +1) : In the party equilibrium, the aggregate surplus W P equals W P = n A ps n A ( +1) n B n C : The next proposition gives a neessary and a su ient ondition for the party equilibrium to reate a larger surplus: Proposition 11 W P W N if and only if ps : Proof. In the appendix. When (i) the number of politiians per party,, (ii) the number of itizens per politiian,, and (iii) the linking osts,, are su iently small vis à vis the itizen s expeted share of the projet, ps, then the intermediation is soially preferred. Note that we have ases when the party equilibrium reates a smaller surplus even if it exists and ases where the party equilibrium reates a larger surplus and, yet, the market agents may oordinate the no party equilibrium. To see this, suppose that the seond inequality in the existene ondition for the party equilibrium, (15) holds, that is ( +) ++. This an be ahieved, for instane, by setting =. The rst inequality in (15) sets a lower bound for ps. This inequality is far less stringent than the ondition of proposition 11. When the latter holds, so does the lower bound for ps ps in (15). Thus, if we hoose a large enough, the party equilibrium exists and generates a larger surplus than the no party equilibrium. Yet, when we hoose ps exists but generates a lower surplus than the no party equilibrium. Thus, depending on the suh that ++ ( +) ps < , the party equilibrium 16

19 parameter values of the model, intervention may be needed in order to enfore the no party equilibrium or the party equilibrium depending on the parameter values of the model. The politiians prefer the no party equilibrium, and they have an inentive to bring about institutions to guarantee its emergene. Sometimes, the interest of the soial planner oinides with the interest of the politiians and promoting the emergene of suh institutions may be in the planner s interest. Yet, when the party equilibrium is preferable, the soial planner should make every e ort to prevent the emergene of suh institutions. Furthermore, even the politiian s individual inentives may render the party equilibrium unstable for instane if the linking osts are su iently low. 7 Empirial Evidene Our framework suggests that the value of belonging to a politial party is inreasing in the value of rents distributed by politiians to itizens. Our simple model assumed that potential party members are idential. With heterogeneous itizens, a member s valuation of belonging to a party should translate into di erent levels of party membership, rather than just to di erent membership fees, osts and personal investments. If the hane to be nominated to a post by a politiian of the party is an important motivation to join politial parties, then the number of itizens willing to pay the osts for joining should be inreasing in the total value of these nominations. 19 The latter an be approximated by the GDP share of total ompensation to publi setor workers. Similarly, if itizens join to seek rents distributed through government ontrats then the publi expenditures share of GDP should be positively related to party membership. As reported in the introdution, there is a strong orrelation between the share of voters belonging to a politial party and the total ompensation osts for the publi setor employees as a perentage of GDP. The following series of regressions on rms this: Table 1 has four regressions for the ountries in Figure 1, exluding Austria. 0 In the rst regression, we use the total ompensation osts for the publi setor (as a perentage of GDP) as an explanatory variable for the share of voters belonging to a politial party. The e et is statistially signifiant at the 10 perent level. In the seond regression, we ontrol for the e et of population size. This is intended to apture the possibility that itizens may be politially more ative in small ountries for reasons outside our model, like greater proximity in smaller ommunities. In the third regression, we use as an additional ontrol the logarithm of GDP per apita evaluated at the purhasing power parity. This ats as a ontrol for di erent levels of eonomi 19 As an important part of the osts of and fees for belonging to a party are non monetary, like providing voluntary work, it is not possible to test our model simply by studying membership fees in di erent ountries. 0 Data from Mair and van Biezen (001), OECD (001), OECD (005) and Statistis Finland (005). 17

20 development. In the fourth regression, we ontrol for both at the same time. TABLE 1. Explaining party membership. Dependent variable: share of voters belonging to a politial party (N=13) Interept (1.677) (6.569)* (44.318) (43.803) Compensation osts (as % of GDP) (0.150)* (0.16) (0.158)** (0.177) Log of population (0.565) (0.590) Log of GDP per apita (PPP) (4.358) (4.146) Adjusted R p value Notes: Standard error in parentheses. * (**) denotes signi ane at the 10% (5%) level. Given the small number of ountries, our results should be interpreted as suggestive, rather than as strit tests of the theory. Nonetheless, our ndings lend some support to our model. Party membership is positively related to the value of rents that politiians distribute, as measured by the ompensation osts of publi setor workers relative to GDP. This relationship is weaker, but does not disappear, when additional ontrols are used. Moreover, we run the regressions using the GDP share of publi expenditures indiating the extent of politially distributed rents, rather than just ompensation osts relative to GDP. The results turned out qualitatively similar, but with statistially weaker e ets. 8 Conlusion In this paper, we suggest a novel explanation for the role of politial parties: party politiians only distribute rents to other party members, thus anelling out the inentive of non members to seek rents and spend resoures in onstruting links to the parties politiians. This redues link formation and enables the party to reate value for the party members. Indeed, our empirial analysis suggests that party membership is more ommon in ountries with a bigger publi setor. To di erentiate the role of publi employment from general inome redistribution, we foused on the ompensation osts of the publi setor employees as perentage of GDP. We take as our starting point anti orruption laws whih forbid itizens from paying politiians diretly for nominations or publily ommissioned projets. These laws still allow itizens (or rms) to pay for gaining aess to politiians, for example by buying tikets to fund raising events, volunteering, and making ampaign ontributions. Competition for politiians attention results in wasteful network formation, whih politial parties may alleviate, when itizens an belong to only one party. Politial parties an redue network formation osts by attahing eah party member to a given politiian, rather than allowing them to build links to several politiians. Similarly, politial parties require politiians to give projets to their party s members. 18

21 It should be highlighted that anti orruption laws also restrit the ativities of politial parties. They are not allowed to trade in nominations or projets diretly, but only to reeive membership payments and alloate funds to politiians ampaigns. Even politial parties are unable to fully eliminate wasteful network formation, as they annot ommit to restriting the number of members. 1 Even though politial parties may save network formation osts, they need not. There are two reasons for this result whih may at rst glane appear ounterintuitive. First, the use of politial parties typially neessitates more formal network struture (there are only two degrees of strengths of the link, either there is a link or there is not). Politiians and itizens, on the other hand, may have more informal links with varying degrees of strength of the link. Seond, maintaining links to the politial parties is ostly. When parties are present, the network must over the parties in addition to politiians and itizens, reating a need for additional links. The use of politial parties improves welfare whenever the linking osts, the number of politiians per party boss and the number of itizens per politiian are su iently low and the expeted rent for the itizen is su iently large. Yet, when the number of itizens per politiian or the number of politiians per party boss is high, but not too high to prevent the party equilibrium from emerging, the no party equilibrium is soially preferred. We also nd that politiians would be better o without politial parties while itizens may gain from their presene, despite the fat that parties harge membership fees from itizens and transfer money to politiians. This surprising result arises as payments by itizens might be even higher without parties. Even so, a party equilibrium may be inentive ompatible in that no single politiian would nd it optimal to deviate from it, provided that others do not. If a party equilibrium results in higher welfare, the problem of the soial planner is to prevent the emergene of politiians olletive bodies that would guarantee the oordination of the no party equilibrium. If the party equilibrium results in higher welfare and is not inentiveompatible, then the soiety may hange this by inreasing the osts of individual politiians to deviate from it. This may explain, for example, why publi money to politial ampaigns is often hannelled through politial parties, rather than diretly to politiians. Our framework raises several topis for further researh. First, we ould endogenize the identity of politiians in the itizen andidate tradition pioneered by Osborne and Slivinski (1996) and Besley and Coate (1997). Seond, we ould endogenize the identity of party bosses by presenting an overlapping generations framework in whih party bosses arise from senior politiians. Finally, Persson and Tabellini (003) show that eletoral rules have signi ant onsequenes on the organization of politial parties and on eonomi poliy. To what extent do these di erenes arise through the role that politial parties play in network formation? 1 Allowing politial parties to pre ommit not to take additional members would disenfranhise those itizens not belonging to the seleted few from fully partiipating in the politial life. 19